Study On The Image Segmentation Algorithm Of Epigraphy Rubbings Based On Fuzzy Clustering

To go on with further analysis and research on the Epigraphy rubbings, the image segmentation is adopted to separate the character or graph with historic and research value from the background for the purpose of making them identifiable and understood, which is significant in the protection of historical relics and cultural heritage. Fuzzy clustering is good at reflecting the uncertainty and fuzziness of the graphs of Epigraphy rubbings. To segment the graph of Epigraphy rubbings with the image segmentation algorithm based on Fuzzy C-Means (FCM) clustering is of better effect than with the traditional segmentation algorithm.The article is about the research on the improvement of FCM clustering algorithm and its application to the image segmentation of Epigraphy Rubbings. The research content and innovation of this paper are mainly indicated on three aspects as follows:(1)As to the uncertainty and fuzziness of the graphs of Epigraphy rubbings, the author analyzes common image segmentation methods and points out the ineffectiveness of prior typical segmentation methods. Fuzzy C-means clustering algorithm is the suitable segmentation method for the uncertainty and fuzziness of the graphs of Epigraphy rubbings.(2)As to the probe on application of standard FCM clustering algorithm in the image segmentation of Epigraphy rubbings, the author analyzes its effect of segmentation on graphs of Epigraphy rubbings with different noise and points out its advantages and disadvantages in application. Experiments proved that the standard FCM is of worse effect on the graphs with much noise when.(3)As to the disadvantages of spending too much time on computation and being ineffective when standard FCM directly applied to the image segmentation of Epigraphy rubbings, the author puts forward an improved fast FCM clustering image segmentation algorithm. To carry out fast clustering, the author categorizes the graphs by the gray scale histogram and its emerging frequency with standard FCM algorithm. On the basis of fast clustering, the membership function is partly improved by dividing the category of graph pixel with the character of pixel neighborhood in the principle of maximum membership. Experiments proved that the improved algorithm gained the advantages of having better ability of suppressing noise, faster in segmentation and computation.